METHODS

All patients admitted to The Hospital for Sick Children (HSC) from 1 January 1996 to 31 March 1998, with a discharge diagnosis of UTI were identified by searching the computerised health records for urinary tract infection, pyelonephritis, and acute pyelonephritis. HSC is a tertiary care children's hospital in Toronto, Canada.

We reviewed the charts of all children under 5 years of age admitted with UTI, defined as a catheter proven culture of a bacterial growth greater than or equal to 10⁵ colony forming units/ml. Children with known underlying genitourinary abnormalities or a history of UTI were excluded.

At HSC, the current standard of care requires that all children under 5 years of age admitted with a culture proven UTI be investigated with ultrasound and VCUG. The timing of these tests is left to the discretion of the attending paediatrician. In this study, a renal ultrasound was considered suggestive of VUR if “dilatation of the pelvi-calyces”, “dilatation of the ureters”, or “dilatation of the collecting system” of one or both kidneys was reported. All ultrasound scans and VCUGs were performed at HSC and the results reported by staff paediatric radiologists. Reflux was graded using the international system of radiographic grading of VUR.⁵

At HSC, voiding cystourethrograms are performed in a standardised manner. The urinary bladder is catheterised with a flexible #8 French silastic catheter without a balloon and taped into position. Hypaque 18% contrast material is then instilled by gravity from a height no more than 1 m above the fluoroscopy tabletop with the patient in a supine position. Intermittent fluoroscopy is performed during the filling stage to detect VUR or other abnormality. A spot image of the filled bladder is normally obtained. Once the bladder is sufficiently filled, infants and young toddlers will spontaneously void. Older children are instructed to void when they feel full, girls lying supine on a bedpan and boys in a left anterior oblique position over a urinal. Spot images are obtained of bladder and urethra during voiding, the catheter having either been pulled out or fallen out. A post-void image of the bladder and renal fossae is obtained to assess post-void volume and reflux. Ultrasound scans were performed with Acuson Sequoia and ATL 5000 equipment. All studies were performed with sector, curved array, and linear high resolution transducers.

Descriptive statistics for the entire group included age, sex, and prevalence of VUR. The statistical indices used to evaluate the properties of the diagnostic test were sensitivity, specificity, and predictive values (positive and negative). When a gold standard test exists (in this case, VCUG), sensitivity is defined as the proportion of those with the disorder (VUR) in whom the test (renal ultrasound) is affirmative; specificity is the proportion of those without the disorder in whom the test is negative. From a clinical perspective, however, these two indices are not as helpful as predictive values. The positive predictive value in this context is the probability of VUR in children with dilatation noted on the ultrasound scan; the negative predictive value is the probability of no reflux in children with no dilatation noted on ultrasound. Unlike sensitivity and specificity, predictive values are influenced by the prevalence of disease. For all indices, 95% confidence intervals were calculated around the point estimates.

RESULTS

A total of 202 charts of patients less than 5 years of age with a discharge diagnosis of UTI were identified. Thirty two were excluded because of a previous episode of UTI or a negative catheter urine culture, leaving 170 who fulfilled the inclusion criteria of a first episode, catheter proven UTI. Eight patients who were booked for a VCUG as outpatients did not return for their appointment, leaving 162 patients with a first episode UTI who had a VCUG performed.

The median age of the 162 subjects was 85 days; 71 (44%) were female. The median age for males was significantly lower than for females (74 v 157 days, respectively, p < 0.05). The age range for males was 6–712 days, and for females 3–1606 days. The median number of days between start of treatment for UTI and ultrasonography was 2 days (interquartile range: 1 day, 3 days). Thirty five patients were found to have VUR on VCUG, giving a prevalence of 22%. Of these 35 patients, two had grade I reflux, 20 had grade II reflux, six had grade III reflux, seven had grade IV reflux, and none had grade V reflux. Ultrasound results suggested the presence of VUR in 14 of these 35 patients, and in 30 of 127 patients without VUR on VCUG. Table 1⇓ shows the distribution of the 35 patients with VUR on VCUG by grade and ultrasound result.

The sensitivity of ultrasound for detection of VUR was 40% (95% CI: 24% to 56%); specificity was 76% (95% CI: 69% to 83%). The positive predictive value of ultrasound for VUR was 32% (95% CI: 18% to 46%) and the negative predictive value was 82% (95% CI: 75% to 89%).

Predictive values were calculated for subjects less than 18 months (151 of the 162 subjects were less than 18 months). The positive predictive value of ultrasound for VUR was 30% (95% CI: 18% to 46%) and the negative predictive value was 82% (95% CI: 73% to 88%). Because only 11 subjects were older than 18 months of age, predictive values were not calculated for this age group.

DISCUSSION

This study shows that ultrasound findings are not predictive of VUR on VCUG. Furthermore, some children with higher grade VUR by VCUG were not detected by renal ultrasound, including five children with grade III and two with grade IV reflux. Thus, in our opinion, a VCUG is necessary to rule out VUR, regardless of renal ultrasound findings.

For a first episode UTI in infants, renal ultrasound is performed to rule out anatomic abnormalities of the urinary tract. It is a non-invasive form of imaging performed before the VCUG and after the initial infection. Various degrees of dilatation of the collecting system of the kidney seen on renal ultrasound are often reported. A Medline search, however, found only three studies that examined the significance of these findings in children. Davey and colleagues⁶ looked at older children (mean age 4.2 years) who were referred for renal ultrasound and VCUG for a variety of indications, including UTI. They found that the frequency of VUR in children with mild renal pelvic distension did not differ significantly from that in children with no distension on renal ultrasound (39% v 32%, p = 0.365). Blane and colleagues⁷ retrospectively analysed VCUG and ultrasound results of 493 children. All children who had a VCUG within eight hours of a renal ultrasound scan were included, except for children with myelomeningocoele or renal surgery. The mean age of their study population was 4.9 years. They found that ultrasound was not sensitive for VUR. Of the kidneys with VUR, 74% had normal ultrasound scans. DiPietro and colleagues⁸ found that ultrasound was unreliable in excluding VUR in children aged 5 years or older who were being evaluated for a UTI. Only two of 21 children with VUR on VCUG had abnormal renal ultrasound scans.

Studies have reported on the reliability of colour flow Doppler sonography in the diagnosis of VUR in children.^9,10 The relation between various degrees of renal collecting system dilatation on fetal ultrasound and postnatal VUR has also been reported; however, this does not apply to our study population.^11,12

The median age of the children in our study was 85 days, reflecting the group that often requires hospitalisation for treatment of UTI. They were considerably younger than the children studied previously.^6–,⁸ The prevalence of VUR in our study population was 22%, an estimate similar to that found in previous studies in the same age group.² The compliance rate for completing the VCUG after the renal ultrasound was excellent (162/170, 95%).

Our study, being retrospective in design, had some limitations. Only children admitted to the hospital were included. All ultrasound and VCUG studies were reviewed by a group of six staff paediatric radiologists. However, inter-rater reliability was not assessed.

Radiologists reading the VCUG results were not blinded to the ultrasound results. Any bias, however, would have been expected to be towards reporting more VUR in patients with “suggestive” ultrasound findings, a bias not evident in the study results.